Experimental Research on Damage Detection Based on Time Domain Data and Bayesian Fusion

Abstract

Abstract Damages such as cracks often occur in engineering structures, and such damages often have the characteristics of time-domain variable stiffness. To effectively identify this kind of damage, a damage detection method based on time domain data and Bayesian fusion is presented in this paper. First, a hybrid model of the AR/ARCH model is used to identify structural nonlinear damage, and a damage indicator is also used: the second-order-variance indicator (SOVI). Although most of the nonlinear damage information is extracted by the indicator SOVI and ARCH model, some nonlinear information is still filtered out by the AR model part. Therefore, a linear cepstral metric indicator (CMI) based on the AR model is introduced to extract the remained nonlinear damage information, and further, the Bayesian fusion theory is applied to combine the results of SOVI and CMI to obtain complete nonlinear damage information and achieve better identification results. Finally, a three-story frame experimental model is used to demonstrate the effectiveness of the Bayesian fusion. Experimental results show that the Bayesian fusion method is superior to SOVI and CMI.